On 01/17/2013 08:24 PM, Mike Burns wrote:
On Thu, 2013-01-17 at 09:43 +0800, Wenyi Gao wrote:
On 2013-01-16 19:47, Mike Burns wrote:
On Wed, 2013-01-16 at 13:33 +0800, Wenyi Gao wrote:
Hey,
Currently, the ovirt-node hypervisor runs in livecd mode, that is the
root filesystem is read-only.
We have to run "persist" command for persistent files, which is
inconvenient for some uses who
run ovirt-node only as hypervisor alone. So do we have plan to let
ovirt-node run non-livecd mode?
Thanks.
Hi Wenyi,
ovirt-node doesn't just run in livecd mode, it *is* a livecd. If you're
using ovirt-node with ovirt-engine (or derivative builds like RHEV-H
with RHEV-M), then you shouldn't have any need to persist anything
manually. This is handled by the vdsm and ovirt-node packages
automatically.
There is no plan to let it run in any mode other than livecd. There are
other options in those cases, like running a full OS rather than an
ovirt-node iso.
Thanks
Mike
Hi Mike,
Thank you very much for answering the question. My case is we are using
ovirt-node
as hypervisor standalone without ovirt-engin or RHEV-M.
So could you have some ideas to keep the rootfs writeable into disks
persistently
not into tmpfs? Like overlay file system?
There isn't really a way unless you are building the image yourself from
scratch, and even then, some things won't work as you would expect them
to.
One possibility:
* add / to the rwtab or an rwtab.d file -- this makes everything
writable, though not persistent. It needs to be done at build
time, though.
* when installing, install with a very large Config partition.
It's 5MB by default, but you will need far more than that.
* persist the directories that you want persisted. This will
automatically persist everything in those directories
to /config. You probably don't want/need /var persisted since
that is generally runtime-only data. /data /var/log and /config
are already backed by real storage, so no need to persist those
either.
In general, I don't recommend this. There may be things that just don't
work the way you would expect. The method for combining the 2
filesystems is done through bind-mounts which happen during the boot
process. Therefore, some things that get persisted don't get
bind-mounted early enough in the process to be used correctly. These
need special handling in the ovirt-early service.
One other thing you could try is altering /etc/fstab in the base image
to symbolically link some directory in /data. /data is backed by real
storage, so that would theoretically work as well and might be safer.
The change needs to be done in the base image at build time or using
something like a plugin to make this change.
Hope this helps
What if we install ISO to disk and not using the way copying the
squashfs img, we copy the squashfs filesystem and mount root fs as rw,
like the way we install a LiveCD Fedora to disk.
Do you think it's possible the ovirt-node would support in case some
custom won't use a stateless ovirt-node and not working with ovirt-engine.
-Simon
Mike
Best regards
Wenyi Gao
_______________________________________________
node-devel mailing list
[email protected]
http://lists.ovirt.org/mailman/listinfo/node-devel
_______________________________________________
node-devel mailing list
[email protected]
http://lists.ovirt.org/mailman/listinfo/node-devel
_______________________________________________
node-devel mailing list
[email protected]
http://lists.ovirt.org/mailman/listinfo/node-devel
